Enabling Fake Base Station Detection through Sample-based Higher Order Noise Statistics

This paper presents computationally efficient fake base station (FBS) detection scheme through higher order statistical analysis at the user equipment (UE) side. In the proposed RF fingerprinting detection scheme, the UE inspects surrounding base stations (BS) by first extracting noise from the received signal through novel sample-based parametric estimation technique and then measuring the noise structuredness with the aid of fourth order moment i.e. kurtosis over the estimated noise samples. This reveals unique RF fingerprints of the legitimized regular base station (RBS) in terms of hardware impairments and resultantly indicates minimal impact from the non-linearities. In contrary, FBS exhibits different RF fingerprints containing larger amount of non-linearities in the received signal due to presence of large impairments. With the help of actual measurement results from RBS and various software defined radio (SDR) based FBS at different cellular standards and by defining a critical threshold of detection, we show that an FBS deviates a lot from the ideal Gaussian noise distribution and constitutes of multivariate distributions, whereas an RBS show minimal deviation from the reference and contains univariate noise distribution. We further calculate the observation sample time required to detect an FBS in an optimal MMSE sense and indicate with the help of results that the minimum time to identify a fake cell is 10ms.